J. Non-Newtonian Fluid Mech.

165 (2010) 13571372

Contents lists available at ScienceDirect
Journal of Non-Newtonian Fluid Mechanics
j our nal homepage: www. el sevi er . com/ l ocat e/ j nnf m
Laminar, transitional and turbulent annular ow of drag-reducing polymer
solutions
A. Japper-Jaafar
a,b
, M.P. Escudier
b
, R.J. Poole
b,
a
Mechanical Engineering Department, Universiti Teknologi Petronas, 31750 Tronoh, Perak, Malaysia
b
Department of Engineering, University of Liverpool, Brownlow Street, Liverpool L69 3GH, Merseyside, United Kingdom
a r t i c l e i n f o
Article history:
Received 20 April 2010
Received in revised form 30 June 2010
Accepted 1 July 2010
Keywords:
Annular ow
LDA
Transition
Shear thinning
Polymer solutions
a b s t r a c t
Mean and rms axial velocity-prole data obtained using laser Doppler anemometry are presented
together with pressure-drop data for the ow through a concentric annulus (radius ratio k =0.506) of
a Newtonian (a glycerinewater mixture) and non-Newtonian uidsa semi-rigid shear-thinning poly-
mer (a xanthan gum) and a polymer known to exhibit a yield stress (carbopol). Awider range of Reynolds
numbers for the transitional ow regime is observed for the more shear-thinning uids. In marked con-
trast to the Newtonian uid, the higher shear stress on the inner wall compared to the outer wall does
not lead to earlier transition for the non-Newtonian uids where more turbulent activity is observed in
the outer wall region. The mean axial velocity proles show a slight shift (5%) of the location of the
maximum velocity towards the outer pipe wall within the transitional regime only for the Newtonian
uid.
2010 Elsevier B.V. All rights reserved.
1. Introduction
The study of uid owing through an annulus has been the sub-
ject of interest since the early work by Rothfus et al. [1] which
concerned a Newtonian uid. Annular ow of a non-Newtonian
uid has a number of engineering applications, especially in the oil
industry where the drilling uid (or mud) is pumped down the
drill string, through the drill bit and nally up the annulus, during
the oil and gas drilling process. There are also other applications
such as in the food, chemical and pharmaceutical industries, that
involve non-Newtonian uids owing through annular pipes [2].
Newtonian and non-Newtonian uid ow in a circular pipe is
symmetrical in the laminar and fully-turbulent regimes [37]. In
such cases the positions of zero shear stress and maximumvelocity
coincide. In the transitional-ow regime, however, while the ow
of Newtonian uids remains symmetrical on average, strong asym-
metry has been observed for a wide range of non-Newtonian uids
[6,812]. In contrast with the pipe-ow situation, fully-developed
annular ow in the laminar and turbulent regimes involves a com-
bination of two boundary layers extending fromthe pipe walls to a
point of maximum velocity which does not lie in the centre of the
annular gap. The two boundary layers are of different thicknesses
and possibly of different ow regimes due to the different degrees

Corresponding author. Tel.: +44 151 7944806; fax: +44 151 7944848.
E-mail addresses: azuraien japper@petronas.com.my (A. Japper-Jaafar),
escudier@liv.ac.uk (M.P. Escudier), robpoole@liv.ac.uk (R.J. Poole).
of curvature of the inner and outer pipe walls [13]. As a result of the
interaction of these two layers the owis then asymmetric. Hence,
in this type of ow, the position of zero shear stress will not, in
general, be coincident with the position of maximum velocity. It is
also the case that turbulent transport phenomena are expected to
differ from those of symmetric ows in circular pipes [1416].
Early investigations for the ow of Newtonian uids by Knud-
sen and Katz [17], Rothfus et al. [1,18], Walker and Rothfus [19]
and Brighton and Jones [20] assumed coincidence of the position of
zeroshear stress andmaximumvelocity. LawnandElliott [21] were
the rst to show, using hot-wire anemometry, that the positions of
zero shear stress and maximum velocity for fully-developed tur-
bulent ow are non-coincident. They found that for radius ratios
k less than 0.4 the position of zero shear stress was closer to the
inner pipe wall thanthat of the maximumvelocity. The inner veloc-
ity proles were also found to deviate from the well-known log
law (derived from pipe-ow data). Their ndings were later sup-
ported by Rehme [14,22] who studied fully-developed turbulent
ow of air using hot wire anemometry in concentric annuli of
varying radius ratios (k =0.02, 0.04 and 0.1). Nouri et al. [15] were
the rst authors to employ laser Doppler anemometry (LDA) for
both Newtonian and non-Newtonian annular-ow measurements
(Re 8900 for Newtonian and Re 1150 for non-Newtonian ows).
They measured the Reynolds shear stress of a Newtonian uid in
an annulus of radius ratio of 0.5 and found that within their exper-
imental uncertainty the location of zero shear stress could not be
distinguished from that of the maximum velocity. These positions
were found to be closer to the inner pipe wall and independent of
0377-0257/$ see front matter 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.jnnfm.2010.07.001
1358 A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372
Nomenclature
a CarreauYasuda parameter
A Cross-sectional area (m
2
)
b Diameter of CaBER plates (4mm)
c Polymer concentration (%, w/w)
D
CaBER
Filament diameter in CaBER (mm)
D
o
Midpoint lament diameter in CaBER following ces-
sation of stretch deformation (mm)
DR Drag reduction (%) (((f
n
f
p
)/f
n
) 100)
f Fanning friction factor ( 2z
W
]U
2
8
)
h Distance between plates in CaBER (mm)
HB HerschelBulkley number (z
y
/K(U
B
/R)
n
)
k Power law consistency index (Pa s
n
)
K HerschelBulkley consistency index (Pa s
n
)
L Pipe length over which the pressure drop was mea-
sured (m)
m Slope of linear tting to CaBER data (mm/s)
n Power-law exponent
p Pressure (Pa)
^p Pressure drop (Pa)
P Wetted perimeter (m)
Q Volumetric ow rate (m
3
/s)
r Radial location (m)
R Pipe radius (m)
r
max
Point of maximum velocity (m)
r
p
Extent of plug region (m)
r
z=0
Radial location of zero shear stress (m)
Re Reynolds number based on viscosity at the wall
(U
B
D
H
/q
W
)
Re
crit
Critical Reynolds number obtained fromtime traces
of the mean axial velocity
Re
1
First Reynolds number limit representing onset of
transition
Re
2
Second Reynolds number limit representing offset
of transition
t Time (s)
T
B
Filament break-up time in CaBER (s)
Tr Trouton ratio

3 )]q( )

u Mean axial velocity (m/s)

u
z
Friction velocity (m/s)

z
W
]

rms axial velocity uctuation (m/s)

U
B
Bulk velocity (m/s) ( Q]((R
2
c
R
2
c
)))
U
local
Local mean velocity (m/s)
U
max
Maximum velocity (m/s)
v

rms radial velocity uctuation (m/s)

w

rms tangential velocity uctuation (m/s)

y Distance from pipe wall (m)
Greek letters
Time ratio
Strain rate determined from CaBER (1/s)
((4/D
o
)(dD
CaBER
/dt))
Shear rate (1/s)
q Shear viscosity (Pa s)
q
E
Uniaxial extensional viscosity (Pa s)
q
o
Zero-shear rate viscosity (Pa s)
q
W
Wall shear viscosity obtained from z
W
and
W
via
CarreauYasuda t (Pa s)
q

Innite-shear rate viscosity (Pa s)

k Radius ratio, (R
i
/R
o
)
z Characteristic relaxation time in CaBER (s)
z
CY
CarreauYasuda constant representing onset of
shear thinning (s)
Fluid density (kg/m
3
)
u

Reynolds shear stress (Pa)

z Stress (Pa)
z
A
Average wall shear stress (Pa)
z
W
Wall shear stress (Pa) (^pD/4L)
z
y
Apparent yield stress (Pa)
Non-dimensional radial location in annulus
((r R
i
)/(R
o
R
i
))
Subscripts
H Hydraulic
i Inner pipe
max Maximum
n Newtonian
o Outer pipe
p Polymer
W Wall
Superscripts
+ wall coordinates
the Reynolds number. In wall coordinates, the inner and outer wall
proles of Newtonianows were foundto obey the well-knownlog
law. However, due to opacity of the non-Newtonian uid used, car-
boxymethylcellulose (CMC), the Reynolds shear stress could not be
measured and therefore maximum velocity and zero shear stress
coincidence was assumed. By performing Direct Numerical Simula-
tions of turbulent concentric annular ow of a Newtonian uid for
radius ratios of 0.1 and 0.5 at Re =8900, Chung et al. [16] conrmed
that the positions of zero shear stress are closer to the inner wall
than those of the maximum velocity for both radius ratios though
in the k =0.1 geometry the effect is more severe.
Escudier et al. [23] performed LDA measurements together
with numerical simulations of a shear-thinning uid, a xanthan
gum/carboxymethylcellulose mixture, in a concentric annulus of
radius ratio 0.506 within the laminar-ow regime. The velocity
proles were again observed to be skewed towards the inner
pipe of the annulus. The calculated velocity distributions using
the power-law uid as a model, were found to be slightly at-
ter with reduced peak velocity levels (when normalised with the
bulk velocity U
B
) when compared to the Newtonian uid ow.
Escudier et al. [24] conducted LDAmeasurements of three different
non-Newtonian uids a xanthan gum, a carboxymethylcellulose
and a laponitecarboxymethylcellulose blend within the lam-
inar, transitional and turbulent annular ow regimes (k =0.506).
Measurements were also conducted on a control Newtonian uid,
a glycerinewater mixture. In turbulent ow, at drag-reduction
levels greater than 35%, an upward shift of the universal velocity
prole was observed with a slope greater than that of the log law.
A slight increase in the peak of the axial rms uctuation compo-
nent, normalized with U
B
, compared to the Newtonian value was
seenwhile the tangential component of uctuationwas distributed
almost uniformlyacross theannular gapandsignicantlybelowthe
Newtonian values.
A number of studies have specically investigated transitional
ow of Newtonian uids in annular ow. Rothfus et al. [1]
undertook such a study using air owing through annuli with two-
different radius ratios (k =0.162and0.650) for amoderateReynolds
number range (1250Re 21,600). Using a pitot tube to measure
the velocities, it was found that the position of maximum velocity
is the same for laminar and turbulent owin both annuli. However,
A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372 1359
Fig. 1. Schematic diagram of the annular pipe-ow loop (ow direction is anti-clockwise).
within the transitional regime the position is shifted closer to the
inner pipe. The extent of the Reynolds-number range for the tran-
sitional regime was found to be a function of the radius ratio and
appeared to be wider than for pipe ow.
Walker and Rothfus [19] also used a pitot tube to study the
behaviour of the radial location of maximum velocity r
max
, with
Reynolds number for water owing in an annulus (k =0.331). The
authors discovered that r
max
in fully-turbulent owcorresponds to
that in laminar ow. During transition the values of r
max
obtained
from the velocity proles start to deviate from the laminar ow
position at Re =650. In the early transitional regime, r
max
appears
to shift towards the inner pipe wall and thenreverses its locationas
the Reynolds number is further increased. The outwardprogression
of r
max
continues past beyond the laminar-owvalue with increas-
ing Reynolds number and stops only when a critical Reynolds
number of 2200 is attained. As Reynolds number increases further
r
max
re-approaches its laminar value.
Hanks and Bonner [13] performed a theoretical analysis on the
stability of laminar Newtonian ow within a concentric annulus.
Their theory predicts that the inner ow region is the least stable
of the two ow regions and will undergo transition to turbulence
while the ow in the outer region remains laminar. Consequently,
the wall shear stress on the inner surface will increase signicantly
due to the change inthe momentum-transport mechanismto a tur-
bulent mode. The increase in wall shear stress leads to a shift in the
radial location of the maximum velocity to a higher value, towards
the outer wall. The radial location of the maximum velocity will
reach a maximum once the outer ow region undergoes transition
to turbulence. Beyondthis critical Reynolds number the radial loca-
tion of maximum velocity will decrease to a value corresponding
to that in turbulent ow where r
max,turbulent
<r
max,laminar
.
Unlike the situation for Newtonian uid ow, there is
a very limited literature of experiment and stability anal-
ysis within the transitional ow regime for non-Newtonian
uids. Amongst the available literature, Escudier et al. [24]
monitored the axial turbulence intensity of Newtonian and non-
Newtonian uid ows (xanthan gum, carboxymethylcellulose and
a laponite/carboxymethylcellulose mixture) at the centre of the
annular gap as a means of identifying the onset of transitional ow.
Althoughdeparturefromthelaminar-owregimewas observedfor
the friction factor, f against the Reynolds number, Re(U
B
D
H
/q
W
)
plot, a sudden increase above the noise level in the normalized
axial turbulence intensity, u

/U
B
, was detected at a slightly lower
Reynolds number than what was observed on the fRe plot for all
the uids studied. Despite the circumferential asymmetry high-
lighted by the authors, the mean axial velocity distribution for
the Newtonian control case showed a slight shift of the location
of maximum velocity towards the outer wall within the transi-
tional regime, a trend which was absent for the non-Newtonian
uid ows.
Gucuyener and Mehmetoglu [25] applied Hanks stability cri-
terion [13] to shear-thinning and yield-stress uids. Using the
modied Reynolds number (based on an equivalent diameter and
characteristic parameters of the ow) the authors found that,
regardless of the uid rheology, two very distinct critical Reynolds
number are obtained in a concentric annulus. They claim that the
inner ow region will always have a lower critical Reynolds num-
ber value compared to the outer ow region. Mishra and Mishra
[26], however, predicted only one critical Reynolds number for the
transition to turbulence using Mishra and Tripathis criterion [27]
for power-law uids. This critical Reynolds number is found to be
an increasing function of the radius ratio. Transition to turbulence
is also predicted to be delayed with increased shear-thinning (i.e.
decreasing power-law index).
What is clear from the foregoing is that there is currently no
complete and detailed experimental data set inthe literature, using
reliable measurements techniques (for example laser-based tech-
niques such as LDA or particle image velocimetry), of the mean
ow and turbulence structure (e.g. u

, w

and the Reynolds shear

stress) for transitional and turbulent ows within an annular pipe
for non-Newtonian uids. To address this decit in the current
study we have conducted detailed LDA measurements in an annu-
lar pipe within the three owregimes (i.e. laminar, transitional and
fully turbulent) with particular attention placed on the transitional
ow region and the drag-reducing behaviour of a semi-rigid poly-
mer solution, xanthan gum. The transitional owregime is studied
by monitoring the axial rms uctuation level at xed radial loca-
tions close to the inner and outer walls ( =0.1 and 0.9) and the
time traces of the mean axial velocity at these locations. Laminar
andtransitional measurements arealsoperformedfor anadditional
polymer, carbopol, which is known to exhibit a yield stress.
2. Experimental arrangement
A 5.81-m long annular-ow facility, essentially a modied ver-
sion of the facility described in detail in Escudier et al. [24],
was utilized as shown in Fig. 1. The test pipe [1] comprised four
1041mm long, one 625mm long and one 718mm long precision-
bore borosilicate glass tubes, with an average internal diameter of
100.40.1mm and wall thickness of 50.1mm. The inner cen-
trebody [2] was made of thin-walled stainless steel tube with
an outside diameter of 50.8mm giving a radius ratio, k =0.506
and a length-to-hydraulic diameter ratio of 117. The thin wall
thickness of the centrebody gave a near neutral buoyancy in a
1360 A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372
Fig. 2. Schematic of the slit module.
water-based solution, minimizing hog and sag; i.e. the possibil-
ity of the centrebody to arch upwards or downwards is reduced.
The centrebody was held in position by a thrust bearing located in
the upstream end and a hydraulic jack [3] tensioned to 3tonnes
axial load on the downstream end of the annulus. The uid was
driven from a 500l capacity stainless steel tank [4] by a positive-
displacement progressive-cavity pump [5] (Mono-type, E101 with
a maximum owrate of 0.025m
3
/s) with three pulsation dampers
[6] to smooth the ow. A Fischer and Porter MAG-SM Series 1000
electromagnetic owmeter (model 10D1) with a maximumcapac-
ity of 0.0333m
3
/s [7] was also incorporated in the experimental
facility. The temperature of the uid was monitored using a plat-
inum resistance thermometer positioned inside the tank with an
accuracy of 0.1

C. During a run the temperature of the uid in

the rig never varied by more than 1

C.
Pressure-drop measurements were conducted using a differ-
ential pressure (transducer GE Druck type LPX9381), between
tappings at 75.8D
H
downstream of the inlet to the pipe test and
117.1D
H
. Velocity proles and Reynolds stress measurements were
conducted using a Dantec Fiberow laser Doppler anemometry
(LDA) system comprising of a 60X10 probe and a Dantec 55X12
beamexpander together with Dantec Burst SpectrumAnalyzer sig-
nal processors (model 57N10 and 57N20). The lens focal length
is 160mm and the measured half angle between the laser beams
is 9.14

which produces a measuring volume with a diameter of

36m and a length of 0.22mm in air. Proles of mean velocity
(u) and rms Reynolds normal stress (u

, v

and w

) were performed
at a location 104.7D
H
downstream of the inlet. Poole [28] has
established that for radius ratios k 0.5, the required development
length for laminar Newtonian owis equal to that of the equivalent
channel ow; e.g. a development length of about 88D
H
is required
for Re =2000. As the development length for turbulent ow is sig-
nicantly lower thanthat for laminar ow[29,30], we consider that
the distance fromthe inlet to the location of measurements is more
than sufcient for the ow to reach fully-developed conditions. A
water-lled at faced optical box [8] installed at the measurement
location was used to minimise the refraction of the laser beams
making it possible toobtaindata closer tothe outer pipe wall where
refraction is most severe due to the curved pipe wall. The velocity
measurements and the respective radial locations were corrected
usingthe ray-tracingmethodoutlinedbyPresti [31]. For some mea-
surements a slit module [11] was also used. This module is shown
in Fig. 2 and was adapted fromthe arrangement rst used by Poggi
et al. [32] and more recently by Escudier et al. [5]. It consists here
of an open cross slit on the outer pipe of the annulus which allows
both pairs of laser beams to have the same optical path without
any refraction thus providing a means of measuring two velocity
components (axial and radial) simultaneously using LDA and per-
mitting the determination of the Reynolds shear stress (u

). A
at-facedbox was constructedaroundthe outer pipe withthe cross
slit to prevent leakage through the slit. The openings of the cross
slit were constructed to be as small as possible to minimise distur-
bance to the ow but sufcient to permit the laser beams to pass
through.
At each location across the annular gap, 10,00030,000 data
samples were collected and processed using a simple ensemble-
averaging method. Processing the data using a transit-time
weighting method [33], to account for velocity-biasing effects,
produced minimal differences (average of 2% in the turbulence
component). A maximum statistical error, for a 95% condence
interval, was less than 0.5% in mean velocity and less than 1.5% in
the turbulence intensity based on the method of Yanta and Smith
[34]. The owrates obtained from integration of the LDA mean
velocity proles were found to be within 1.5% of the value provided
by the owmeter.
3. Working uid preparation and characterisation
As Newtonian control uids, we used both water and a
glycerinewater solution. The latter was prepared by rst adding
the 60% (weight) glycerine into the 500l tank. Water was then
added until the total volume of uid was sufcient to fully ll the
rig (500l) during pump operation. The solution was circulated
aroundthe whole owloopuntil homogeneity was achievedwhich
was determined by measuring the viscosity of the uid collected
from the ow loop every 30min. The nal glycerinewater mix-
ture had a density of 1070kg/m
3
and a shear viscosity, measured
at 20

C, of 38.6mPa s.
Therst of thetwopolymers usedherewas xanthangum(Ketrol
TF) obtained in powder form from Kelco Co. with the molecular
weight of an individual xanthan gum chain reported by the sup-
plier to be in excess of 10
6
g/mol. A quantity of approximately
700l of tap water was used as the solvent for the polymer. Prior to
the addition of polymer powder, water was circulated within the
facility to remove any dissolved air. Mixing of part of the solvent
with the powder was achieved by circulating the polymer solution
within the mixing loop [9] at a low pump speed for at least 5h
before the mixing loop was opened and the solution circulated in
the ow loop, allowing further mixing with the rest of the solvent
in the pipe for at least another 5h, until the solutions were visibly
homogeneous. The homogeneity of each solution was also checked
by comparing the viscometric data with a small sample solution
(500ml) of the same concentrationpreparedseparately. To retard
bacteriological degradation, 37%(w/w) formaldehyde was addedto
the polymer solutions at a concentration of 100ppm. The polymer
solution was left to hydrate in the rig for at least 24h prior to the
LDA measurements.
Steady-shear measurements were conducted at 20

C on small
sample solutions (500ml) of xanthan gum prepared separately
outside the ow loop for concentrations in the range 0.010.75%
A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372 1361
Fig. 3. (a) Viscometric data for xanthan gum solutions together with the
CarreauYasuda ts. (b) Zero-shear viscosity versus concentration for xanthan gum
at 20

C.
(w/w) using a TA-Instruments Rheolyst AR1000Ncontrolled-stress
rheometer. For low concentrations (0.1%) a 41.2mm mean diam-
eter double concentric cylinder was used while for the higher
concentrations a 40-mm, 2

cone-and-plate geometry were uti-

lized. The plot of shear viscosity against shear rate for xanthan
gum in Fig. 3(a) shows increased dependence of the shear viscos-
ity, q on shear rate, with increasing concentration i.e. increased
shear-thinning. The CarreauYasuda model [35] was used to t the
data:
q q
c
q
c
q

=
1
[1 +(z
CY
)
u
]
n]u
(1)
where q
o
and q

are the viscosities in the zero-shear and innite-

shear plateaus while z
CY
, n and a are constants which, respectively,
represent the inverse shear rate at the onset of shear thinning, the
power-law index and a tting parameter introduced by Yasuda et
al. The ts were achievedusing the methodology outlinedinEscud-
ier et al. [36] andthe parameter values are providedinTable 1. From
Fig. 3(b), a log-log plot of zero-shear viscosity versus concentration
for xanthan gum, the slope was found to be about 1.4 for the semi-
dilutenon-entangledregion(i.e. belowa critical concentration) and
5.2 for the entangled region. These values agree well with values
available in the literature [3739]. The location of the critical con-
centration for the formation of a polymeric network was found to
be about 0.067%.
Although only possible for concentrations higher than those
used in the uid-dynamic measurements (c 0.2%, w/w), exten-
sional property measurements were possible for xanthan gumand
were carried out using a Capillary Break-up Extensional Rheometer
(CaBER) supplied by Thermo Electron GmbH in conjunction with
a high-speed camera. The CaBER utilizes a laser micrometer, with
a resolution of around 10m, to monitor the diameter of the thin-
ning elongated lament, which evolves under the action of viscous,
inertia, gravitational and elastocapillary forces. High-speed digital
imaging of the process was captured by a Dantec Dynamics Nano
Sense MKIII high-speed camera with a Nikon 60mmf/2.8 lens at
2000 frames per second.
A sample of about 25mm
3
was loaded using a syringe between
the 4mm plates of the instrument, making sure that it was totally
homogeneous withnobubbles withinthe sample. The initial aspect
ratio (h/b) of 0.5 was chosen based on the recommendation by
Rodd et al. [40] to minimize the effects of reverse squeeze ow
and sagging. A uniaxial step strain was then applied, resulting in
the formation of an elongated lament. A linear stretching defor-
mation was employed as the mode of the step strain. The stretch
time was set to 50ms. The nal aspect ratio was varied with solu-
tion concentration such that lament thinning was still observed
between the 4mm plates. For example, a nal aspect ratio of 2.2
was chosen for 0.2% XG in order to observe lament thinning over
a timeframe of about 25ms.
Fig. 4(a) shows the decay of the lament diameter against time
for 0.2% XG which was tted to an equation of the form:
D
CaBER
= D
c
c
t]3z
(2)
as recommended by Stelter and Brenn [41] for viscoelastic u-
ids, where D
o
is the midpoint diameter following cessation of
the stretch deformation and z is a characteristic relaxation time
whichrepresents the characteristic time scale for viscoelastic stress
growth in uniaxial elongational ow[40]. Alinear tting character-
istic of Newtonian-like thinning:
D
CaBER
= mt +D
c
(3)
was also t to the data. The Trouton ratio was calculated using the
equation recommended by Pelletier et al. [42]:
1r =
q

q( )
(4)
with the strain rate calculated from:
=
4
D
c
dD
CaBER
dt
(5)
Newtonian linear thinning was observed over the entire time to
breakup with a slope of about 40mm/s. Linear tting the last six
data points prior to breakup, where the sample most closely resem-
bles the uniformcylindrical shape used in deriving Eqs. (2) and (3),
gave a slope within 10% of the global t. Linear thinning was also
observed for higher concentrations. Despite this apparent Newto-
nian behaviour, the Trouton ratio plotted against concentration in
Fig. 4(b) conrms the non-Newtonian behaviour of this xanthan
gum as the magnitude for all the concentrations studied was sig-
nicantly greater than that of a Newtonian uid, i.e. Tr 3.
The second polymer used here was Carbopol 980 supplied by
Noveon, France, in white occulated powder form with a molecu-
lar weight of 4.0010
6
g/mol. It is a non-toxic version of Carbopol
940 [43,44] and has been reported to showan apparent yield stress
[43] when suitably neutralized. In addition to the procedure listed
above for xanthan gum, therefore, the Carbopol solution was neu-
tralizedduringmixingusinglaboratorygrade2Nsodiumhydroxide
supplied by BDHLtd, UK. Due to the thixotropic nature of carbopol,
the solution within the owloop was circulated at 30% of the maxi-
mumpumpspeedfor 30minprior to any measurements (including
1362 A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372
Fig. 4. (a) Filament diameter thinning as a function of time for 0.2% XG at 20

C
(arrowlines represent the time period for exponential t ( ) and linear t ()).
(b) Trouton ratio versus XG concentration. The error bars represent data variations
calculated from at least four measurements.
Fig. 5. Shear viscosity versus shear stress for carbopol solutions. Hollow symbols:
measured using 40-mmroughened plate, lled symbols: measured using 40mm, 2

cone and plate.

rheological tests). This pre-shear was to ensure reproducibility of
the results by standardization of the solution shear history.
Complete rheological characterization of the carbopol solutions
was not conducted due to the well known difculties associated
with these solutions, such as thixotropy and slip, as outlined by
Presti [31]. Steady-shear measurements were only performed on
0.065% and 0.1% carbopol solutions which were prepared in the
annular rig for the detailed uid dynamic measurements. Due
to wall-slip artefacts [43] which are known to exist for thicken-
ers such as carbopol, a 40-mm stainless steel roughened parallel
plate geometry was used to characterise these solutions in addi-
tion to the 40-mm, 2

cone-and-plate geometry. Fig. 5 shows

the shear viscosity data against shear stress for 0.065% and 0.1%
carbopol solutions measured using the cone and plate and the
roughened-plate geometries. The viscosity data shows clearly that
slip problems, which are most severe at lowshear stresses, exist on
the smooth surface of the cone-and-plate geometry which resulted
in a lower apparent viscosity being measured. For the data mea-
suredusing the roughenedplate, where the slipeffect is minimized,
a very highrst Newtonianregionis observedat lowshear stresses.
Even though a real yield stress may not exist for these solutions,
the abrupt viscosity transition from very high values to low val-
ues within a narrow range of shear stresses is an indication that
a critical shear stress does exist above which the uid exhibits
more liquid-like behaviour. The viscosity data was tted to the
Table 1
CarreauYasuda parameters for xanthan gum solutions.
Concentration
c (%, w/w)
Zero-shear
viscosity qo (Pa s)
Innite-shear
viscosity q (Pa s)
Constant which represents
onset of shear thinning z
CY
(s)
Power-law
exponent n
CarreauYasuda
parameter a
0.01 1.7310
3
1.1210
3
4.9310
2
0.21 29.50
0.0124 3.1510
3
1.1210
3
9.5310
2
0.33 10.00
0.025 6.0210
3
1.1210
3
1.5310
1
0.35 4.98
0.0375 9.6710
3
1.1210
3
2.5210
1
0.37 3.35
0.05 1.9710
2
1.1310
3
3.8810
1
0.42 2.10
0.07 5.3310
2
1.3310
3
8.3410
1
0.49 1.12
0.10 2.2510
1
1.8810
3
1.77 0.60 0.57
0.15 2.16 2.6710
3
2.88 0.80 0.31
0.20 3.68 2.2410
3
2.1510
1
0.66 0.81
0.25 9.83 3.4510
3
2.3410
1
0.75 0.80
0.375 2.8510
2
4.6010
3
4.1410
2
0.81 1.47
0.5 8.4110
2
5.2510
3
7.1110
2
0.84 1.22
0.75 1.5510
3
9.5810
3
1.0410
3
0.85 0.92
A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372 1363
Fig. 6. fRe data for water and glycerine with u

/U
local
levels to monitor transition
for glycerine.
HerschelBulkley model:
z = z
y
+K
n
(6)
where z
y
, K and n are, respectively, the apparent yield stress,
HerschelBulkley constant and the power-law exponent. The ts
are included inFig. 5 and the tting parameters are listed inTable 2.
Measurements of uidrheologyfor all solutions wereconducted
prior to and after each LDA prole traverse to check for signs of
mechanical and bacteriological degradation where a decrease of
more than 5% from the initial state was taken as a sign of degra-
dation and the uid disposed of. The rst normal-stress difference
was found to be belowthe sensitivity of the rheometer for all poly-
mer solutions which indicates that the uids studied are probably
only very weakly elastic in the non-linear regime, at least for the
range of concentrations used here. Seeding particles (Timiron MP-
1005, mean diameter approximately 5m, supplied by S. Blanck
Ltd) were addedat a concentrationof 1ppminorder to increase the
signal-to-noise ratio and the data rate for the LDA measurements.
4. Pressure-drop measurements and transition
identication
Fig. 6 is a plot of the Fanning friction factor, f against Reynolds
number, Re for water and glycerine. Good agreement is observed
with the theoretical prediction for the laminar-ow regime given
by Shah and London [45] for radius ratio k =0.506:
j =
23.82
Rc
(7)
where Re is based on the hydraulic diameter. In the high Reynolds
number, turbulent-ow regime, the data agrees well with the
empirical equation given by Jones and Leung [46]:
1

j
= 4log(1.343Rc j
1]2
) 1.6. (8)
Due to its low viscosity, laminar-ow conditions were not attain-
ablewithintheoperatingrangeof theowloopfor water. However,
the data for glycerine shown encompasses all three ow regimes,
laminar, transition and turbulent. Reasonably good repeatability
was observed for the measurements with an average percentage
difference of 3.7% in the friction factor.
A method initially suggested by Zakin et al. [47] and subse-
quently adopted by Park et al. [48] was used to detect transition
from laminar to turbulent ow by plotting turbulent intensities
measured at xed radial locations against the Reynolds number.
This method is especially useful for the owof non-Newtonian u-
ids where transition is not always well dened in the fRe plot. For
circular pipe ows, Escudier and Presti [8] proposed that the near-
wall (r/R=0.8) axial rms uctuation level is a reliable indicator of
owregime. For this study, the axial rms uctuationlevel was mon-
itored at ((rR
i
)/(R
o
R
i
)) values of 0.1 (inner-wall vicinity) and
0.9 (outer-wall vicinity). In Fig. 6, a plot of the axial rms uctua-
tion component normalized with the local mean velocity against
the Reynolds number for glycerine, a clear demarcation from the
laminar regime can be detected using this method where abrupt
increases in the values are observed from a level of about 3% up
to 22% of the local mean velocity. The low level axial rms uc-
tuations (u

/U
local
<3%) detected within the laminar regime are a
consequence of the combined noise in the LDAsystemand the ow
loop. Normalization by the local velocity is chosen as opposed to
the conventional bulk velocity normalization due to the asymmet-
rical nature of the velocity prole within the annular pipe. The rst
Reynolds number limit, Re
1
, identies the onset of transition seen
as a noticeable increase in the turbulent activity while the second
Reynolds number limit, Re
2
, identies the offset (end) of transition
taken as being where the maximum value of turbulent intensity is
reached (all value all given in Table 3). For this study Re
1
is taken as
theReynolds number wheretheaxial rms levels exceedthe3%noise
level. For glycerine, the rst limit is detected at the same Reynolds
number of 2100 for both the inner and the outer wall while the
second limit is reached earlier for the inner wall at Re 2900 while
for the outer wall Re
2
is about 3100. Note that transition to turbu-
Table 2
HerschelBulkley parameters for carbopol solutions.
Concentration c (%, w/w) Rheometer geometry Yield stress zy (Pa) Consistency index K (Pa s
n
) Power-law exponent n
0.065
Roughened plate 3.8810
1
1.1010
2
0.98
Cone and plate 4.6410
2
3.3210
2
0.81
0.1
Roughened plate 1.14 6.1910
1
0.52
Cone and plate 8.5210
2
1.03 0.44
Table 3
Reynolds number limits and peak axial rms uctuation level.
Fluid Re
1
Re
2
u

max
]U
local
Re
crit
(%)
0.1 0.9 0.1 0.9 0.1 0.9 0.1 0.9
Glycerinewater 2100 2100 2900 3100 0.20 0.22 2300 26 11
0.07% XG 3700 3700 7700 7700 0.15 0.20 6000 15 35
0.15% XG 2400 2400 14,000 11,600 0.13 0.18 6500 23 31
0.065% CARB 2000 2000 3800 3800 0.19 0.26 2800 8 15
1364 A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372
Fig. 7. fRe and u

/U
local
Re data for 0.0124% XG.
Fig. 8. fRe and u

/U
local
Re data for 0.07% XG.
lence for these ows was allowed to occur naturally due to minor
imperfections of the ow loop and uctuations from the pump.
Fig. 7 shows the friction factor data for 0.0124% xanthan gum
where within the ow loop operating range laminar ow was
barely achievable. However at higher concentrations, 0.07% and
0.15% xanthan gum, all three ow regimes, laminar, transition and
turbulent were clearly achieved, as shown in Figs. 8 and 9. The fRe
andu

/U
local
Re datafor 0.065%carbopol showninFig. 10alsoincor-
porate all three ow regimes while for 0.1% carbopol, shown in
Fig. 11, the data spans only the laminar-ow regime. As was the
case for the glycerinewater uid ow, the value of Re
1
is the same
at the inner and outer walls: 3700 for 0.07% xanthan gum, 2400
for 0.15% xanthan gum and 2000 for 0.065% carbopol. The offset of
transition, seenas the maximumaxial rms value, was also observed
at the same Reynolds number for the inner and outer walls except
for 0.15% xanthan gum. In Figs. 810, Re
2
values were found to be
higher for the more shear-thinning uids indicating later transi-
tion offset and subsequently a wider Reynolds number range for
the transitional ow regime. Table 3 lists all the Reynolds number
limits as determined from the near-wall turbulent-intensity mea-
surements and also the peak values of the turbulence intensities for
all uids. It is interesting to note that these peak values decrease
with increasing shear-thinning.
For the Newtonian and polymer ows, time traces of the axial
velocity at locations of 0.1 and 0.9 for various Reynolds numbers
Fig. 9. fRe and u

/U
local
Re data for 0.15% XG.
Fig. 10. fRe and u

/U
local
Re data for 0.065% CARBOPOL.
Fig. 11. fRe and u

/U
local
Re data for 0.1% CARBOPOL.
spanning the three ow regimes were also monitored (although
not measured at the two locations simultaneously). Fig. 12 shows
the time traces for the Newtonian uid, glycerinewater. The plots
show that the ow is completely laminar for Re =2000 and at
Re =2300 spikes are detected at both locations. The velocity spikes
seen from the time traces are not typical of the puffs and slugs
A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372 1365
Fig. 12. Time series of the axial velocity at =0.1 and 0.9 for glycerinewater.
observed in pipe ow by Wygnanski and Champagne [49] and
Rubin et al. [50]. The signals appear to be more complex with com-
binations of high- and low-amplitude spikes. The number of spikes
further increases as the Reynolds number is increasedto2600. Even
higher increases in Re results in more complex behaviour resem-
bling that normally observed in fully developed turbulent ow.
In order to attempt to quantify the degree of turbulence at the
onset of the transitionregiona method knownas the u method [51]
is utilized where a time ratio is dened such that:
=
^t
turbulent
^t
total
% (9)
where ^t
turbulent
is taken as the total time for which spikes occur
in the time trace plot over a period ^t
total
. A spike is considered to
have occurred within the time trace if the peak velocity is differ-
ent by more than 15% of the local mean velocity, U
local
. Within the
laminar regime, e.g. at Re =2000, the velocity varies within 15% of
the U
local
but no obvious spikes are observed.
At theonset of transitionfor glycerineat Re =2300, thetimeratio
for the inner wall was found to be =26% which is higher than that
of the outer wall (=11%). As the time ratio represents the inter-
mittency within the ow, it can be regarded as a direct measure
of how transitional the ow is. Hence, for the Newtonian uid, the
inner-wall owhas a higher degree of turbulence thanthat near the
outer wall at the start of transition. This observationis inagreement
with the behaviour predicted by the theoretical stability analysis
by Hanks and Bonner [13]. At higher Reynolds numbers, the occur-
rence of spikes is more frequent with higher velocity uctuations;
the assessment of the time ratio becomes somewhat subjective and
subject to a high degree of uncertainty.
Fig. 13 shows the time traces for 0.07% xanthan gum at vari-
ous Reynolds numbers. The velocity data at Re =4300 is essentially
steady indicating laminar ow. At Re =6000, the ow close to the
inner wall is clearly unsteady but with no distinct spikes whereas
high-amplitude (>50% of local mean velocity) spikes are seen for
the ow close to the outer wall. This inner wall unsteadiness could
be a consequence of the high amplitude spikes occurring for the
ow closer to the outer wall, i.e. mass conservation enforcing
changes in velocity near the inner wall but the ow remains lami-
nar. The values are 15% and 35% for the inner and the outer walls,
respectively, indicating more turbulent activity near the outer wall.
This observation suggests that the ow close to the outer wall
becomes transitional earlier than the owclose to the inner wall in
markedcontrast tothe corresponding Newtoniancase. The number
of high amplitude spikes increases with Reynolds number and at
Re =28,700 the ow is fully turbulent. Similar characteristics were
observed for 0.15% xanthan gum and 0.065% carbopol (not shown)
with the rst trace of turbulence found at Re
crit
=6500 and values
of 23% and 31% on the inner and outer walls for 0.15% xanthan gum
and Re
crit
=2800 for 0.065% carbopol with time ratios of 8% for the
inner wall and 15% for the outer wall (see Jaafar [52] for complete
details). Table 3 lists the critical Reynolds numbers obtained from
the time traces and the time ratios of the ow close to the inner
and outer wall at the critical Reynolds numbers.
5. Mean axial velocity measurements
Fig. 14shows the meanaxial velocity prole for glycerinewater
at several Reynolds numbers spanning the three ow regimes. The
velocity prole within the laminar regime at Re =2000 is in good
1366 A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372
Fig. 13. Time series of the axial velocity at =0.1 and 0.9 for 0.07% XG.
agreement with the theoretical prole. Even though spikes were
detected in the time trace at Re =2300, the mean velocity prole
remained essentially unchanged and agrees well with the theoret-
ical laminar prole. At both Reynolds numbers, the velocity ratios
of the maximumvelocity to the bulk velocity are within the exper-
imental uncertainty with the theoretical value of 1.51. Deviations
fromthetheoretical laminar proleareobservedat Re =2600where
a slight shift towards the outer wall in the location of maximum
Fig. 14. Velocity proles for glycerinewater at different Reynolds numbers within
the laminar-, transitional- and turbulent-owregimes, including the analytical pro-
le for laminar ow of a Newtonian uid (continuous lines).
velocity is also observed consistent with the higher number of tur-
bulence spots observed at this Reynolds number. As the Reynolds
number is further increased, the shape of the velocity proles tends
to what would be expected of turbulent ow: a progressively at-
ter central region with an increase of the velocity gradient near the
walls. The ratios of the maximum velocity to the bulk velocity at
these Reynolds numbers are 1.28 (Re =2900), 1.24 (Re =3700) and
1.15 (Re =9600) respectively.
For the 0.07% xanthan gum solution the mean axial velocity
prole within the laminar, transitional and turbulent regimes is
presented in Fig. 15. The numerical solutions for laminar ow of a
power-law uid with n=0.61, obtained using Fluent v6.1.22, are
also included for comparison. As expected for a shear-thinning
uid, the ratio of maximum velocity to bulk velocity within the
laminar-ow regime (U
max
/U
B
<1.46) is lower than the theoretical
value for a Newtonianuid(U
max
/U
B
=1.51). Fairly goodagreement
within the laminar-ow regime is obtained with the numerical
results even at Re =6700 where spikes are present in the time trace
plot (Fig. 13). However, due to the at nature of the velocity proles
within the three ow regimes, it is difcult to determine the exact
location of the maximum velocity. At the higher concentration of
xanthan gum (0.15%), Fig. 16, the mean velocity prole coincides
with that of the numerical data with n=0.45 even up to Re =8100
when spikes are again present in the time trace. At Re =15,000
whereindividual spikes couldnolonger bedistinguished, thecalcu-
latedprole for laminar owfailedto predict accurately the prole.
The lower value of n for this uidindicates greater shear thinning of
the higher concentration solution which results in atter proles,
i.e. a slightly smaller ratio of maximum velocity to bulk velocity,
when compared to 0.07% xanthan gum.
A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372 1367
Fig. 15. Velocity proles for 0.07% XG at different Reynolds numbers within the
laminar-, transitional- and turbulent-ow regimes, including the numerical data
from Fluent for laminar ow of a power-law uid with n=0.61 (continuous lines).
Fig. 16. Velocity proles for 0.15% XG at different Reynolds numbers within the
laminar-, transitional- and turbulent-ow regimes, including the numerical data
from Fluent for laminar ow of a power-law uid with n=0.45 (continuous lines).
As the yield stress is low for 0.065% carbopol, no plug region
typical of a yield-stress uid is observed for the laminar-ow pro-
le, and therefore for conciseness, the data is not included here.
However we found better agreement with numerical data for a
Herschel-Bulkley uid with the parameters obtained fromthe data
measured in the rheometer using the smooth cone and plate geom-
etry. If the parameters fromthe data measuredusingthe roughened
plate are used in the numerical simulations, poorer agreement is
observed with the experimental data. These results would seem to
indicate that within the annular owloop wall depletion, i.e. a thin
layer of essentially solvent near the wall, of the carbopol solution
does indeed take place. The lowyield stress has no signicant effect
on the transition to turbulence as the behaviour is similar to that
of the non yield-stress shear-thinning uids. The location of max-
imum velocity within the laminar and transitional ow regimes
was found to be approximately =0.44. However, as the Reynolds
number approaches the fully-turbulent regime, the velocity prole
becomes progressively atter and it becomes difcult to determine
if the location of maximum velocity has in fact shifted. For 0.1%
carbopol, the turbulent-ow regime was not attainable within the
Fig. 17. Velocity proles for 0.1% CARBOPOL at different Reynolds numbers within
the laminar-ow regime, including the numerical data from Fluent for ow of
a HerschelBulkley uid (continuous lines). The non dimensional plug width
rp/(Ro R
i
) =0.25, 0.16 and 0.11 respectively for Re =10, 100 and 4200.
operating range of the ow loop. The laminar proles as shown
in Fig. 17 agree with the numerical results of the Herschel Bulkley
model using the smooth cone and plate parameters suggesting that
wall depletion also occurs in this ow. As can be seen in Fig. 17, the
existence of the plug zone around the central region of the annu-
lar gap is more pronounced for this solution when compared to
0.065% carbopol within the laminar-ow regime as expected from
the fact that the yield stress z
y
is higher for 0.1% carbopol than for
the 0.065% solution. As the Reynolds number is further increased
within the laminar regime, the extent of the plug region reduces
due to the higher viscous shear stresses [10], with the proles also
becoming less at with the ratio of maximum velocity to the bulk
velocity found to be increasing with Reynolds number.
6. Mean ow and turbulence statistics for fully-turbulent
ow
Axial-velocity proles for water together with the Reynolds
normal-stress components have been measured for three differ-
ent Reynolds numbers using the slit module positioned 96.6 D
H
downstream of the pipe entrance. As we discuss below, the use
of the slit module for Newtonian uid ow at Reynolds numbers
greater or equal to 30,000 was successful but for lower Reynolds
number Newtonian-uid ow and for all the non-Newtonian uid
ows, the slit module was found to have an unacceptably large
effect on the turbulent velocity eld and could not be used. The
mean axial velocity proles normalized with the bulk velocity for
the three Reynolds numbers are plotted in Fig. 18, together with
the measurement at Re =9900 where no reliable measurement of
Reynolds shear stress could be made using the slit module. Data
obtained by Nouri et al. [15] for a concentric annulus with radius
ratio0.5for aNewtonianuidat Re =26,600arealsoshownfor com-
parison. Note that the data of Nouri et al. [15] has been rescaled by
6% to match the bulk velocity from the owrate obtained through
the process of integration of the velocity proles as shown to
be necessary by Chung et al. [16]. Flatter proles are observed
with increasing Reynolds number with the location of maximum
velocity found to be located closer to the inner pipe wall at a
non-dimensional location of =0.44 for all Reynolds numbers. This
location is displaced by less than 1% from the location of the zero
shear stress, which is within the experimental uncertainty of the
determination of both locations, as determined from the Reynolds
1368 A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372
Fig. 18. Normalized mean velocity distribution at various Reynolds numbers for
water.
Fig. 19. Reynolds shear stress values normalized with the bulk velocity for water
at various Reynolds numbers. A dotted line is included to showthe theoretical total
shear stress assuming the same shear stress on the inner and outer walls.
shear-stress data plotted in Fig. 19. Since the location of maximum
velocity cannot be distinguished from the location of zero shear
stress, r
max
is taken as both the location of maximum velocity and
zero stress for all the uids investigated here, as Nouri et al. did
for their CMC results, where measurements of the Reynolds shear
stresses were not possible.
Fig. 20 shows the mean ow data of water in wall coordinates
(i.e. u
+
(u/u
z
) against y
+
(yu
z
/q
W
) for all Reynolds numbers. The
friction velocity, u
z
was calculated using the wall shear stresses at
the inner andouter walls. These wall shear stresses were calculated
using the pressure-drop measurements and the zero shear stress
location, in our case we assume r
z=0
=r
max
, using the equations
below [17,15]:
z
c
=

^p
L

R
2
c
r
2
z=0
2R
c

z
i
=

^p
L

r
2
z=0
R
2
i
2R
i
. (10)
The limited data close to the walls shown in Fig. 20 are in good
agreement with that expected for the viscous sublayer (i.e. y
+
<10,
u
+
=y
+
). In the turbulent core region, the data for the inner and the
Fig. 20. Mean velocity distribution for water in wall coordinates. () u
+
=y
+
; (- - -)
u
+
=2.5lny
+
+5.5; (- -) u
+
=2.5lny
+
+4.9 (Clauser); () inner wall, Re =9900; ()
outer wall, Re =9900; () inner wall, Re =30,600; () outer wall, Re =30,600; ()
inner wall, Re =61,400; () outer wall, Re =61,400; () Inner wall, Re =77,000; (v)
outer wall, Re =77,000.
Fig. 21. Mean velocity distribution for 0.0124% XG in wall coordinates. () u
+
=y
+
;
(- - -) u
+
=2.5lny
+
+5.5; (- -) u
+
=2.5lny
+
+4.9 (Clauser); () 0.0124% inner wall,
Re =10,600, DR=3.2%; () 0.0124% outer wall, Re =10,600, DR=3.2%; () 0.0124%
inner wall, Re =30,300, DR=10.3%; () 0.0124% outer wall, Re =30,300, DR=10.3%;
() 0.0124% inner wall, Re =57,600, DR=11.6%; () 0.0124% Outer wall, Re =57,600,
DR=11.6%.
outer walls collapse, in agreement with the log law equation with
the constant proposed by Clauser [53], as opposed to the log-law
constant which is applicable to circular pipe ows [54].
Turbulent-intensity measurements were also conducted for
0.0124% and 0.07% xanthan gum. As has been established in many
previous studies [31,55,56,7], amongst others, the effects of drag
reduction are mainly observed in the regions close to the walls
where differences are observed in peak magnitudes and locations
of the uctuation components in relation to those for a Newtonian
uid. As a consequence of these near-wall effects, in what follows
wall coordinates are used to highlight this importance.
Fig. 21shows the meanowdata inwall coordinates for 0.0124%
xanthangum. The xanthangumdata inthe viscous sublayer follows
u
+
=y
+
as expected. Inthe turbulent core region, the data are shifted
upward frombut remain parallel to the line representing data for a
Newtonian uid. The upshift is clear evidence of drag reduction. At
the lowest measured Reynolds number, where the drag reduction
A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372 1369
Fig. 22. Meanvelocitydistributionfor 0.07%XGinwall coordinates. () u
+
=y
+
; (- - -)
u
+
=2.5lny
+
+5.5; (- -) u
+
=2.5lny
+
+4.9(Clauser); () 0.07%inner wall, Re =28,700,
DR=42.3%; () 0.07% outer wall, Re =28,700, DR=42.3%.
Fig. 23. Axial rms uctuation levels in wall coordinates for water and
0.0124% XG. (outer wall data upshifted on u
+
axis by 1 wall unit). ( )
Water (inner), Re =9900, u
+
max
= 2.44; ( ) water (outer), Re =9900, u
+
max
= 2.63;
( ) water (inner), Re =30,600, u
+
max
= 2.53; ( ) water (outer), Re =30,600, u
+
max
=
2.86; ( ) water (inner), Re =61,400, u
+
max
= 2.68; ( ) water (outer), Re =61,400,
u
+
max
= 3.21; () 0.0124% XG (inner), Re =10,600, u
+
max
= 2.57; () 0.0124% XG
(outer), Re =10,600, u
+
max
= 2.41; () 0.0124% XG (inner), Re =30,300, u
+
max
=
2.63; () 0.0124% XG (outer), Re =30,300, u
+
max
= 2.63; () 0.0124% XG (inner),
Re =57,600, u
+
max
= 2.78; () 0.0124% XG (outer), Re =57,600, u
+
max
= 2.84. (For
interpretation of the references to colour in this gure legend, the reader is referred
to the web version of the article.)
is only about 3.2%, the outer wall data lie close to the line for a New-
tonian uid and are progressively upshifted with higher Reynolds
number. Complete collapse of the inner- and outer-wall data only
occurs at DR12%. These observations suggest that the initial con-
tribution of drag reduction comes from the inner wall where the
wall shear stress is higher than on the outer wall. At even higher
levels of drag reduction, seeninFig. 22 for 0.07%xanthangummea-
sured at Re =28,700 where DR=42.3%, both the inner and outer
walls data are no longer parallel shifted from the Newtonian line,
as expected for high drag-reducing ows [55]. The outer wall pro-
le is lower than the inner wall data indicating that the inner wall
owcontributes more to the overall drag-reduction effect (assum-
ing, of course, that the inner:outer split of pressure drop implied
by Eq. (10) is correct).
Fig. 24. Axial rms uctuation levels in wall coordinates for water and 0.07% XG.
(outer wall data upshifted on u
+
axis by 1 wall unit). ( ) Water (inner), Re =30,600,
u
+
max
= 2.53; ( ) water (outer), Re =30,600, u
+
max
= 2.86; () 0.07% XG (inner),
Re =28,700, u
+
max
= 3.16; () 0.07% XG (outer), Re =28,700, u
+
max
= 2.96. (For inter-
pretation of the references to colour in this gure legend, the reader is referred to
the web version of the article.)
Fig. 25. Radial rms uctuation levels in wall coordinates for water and 0.0124%
XG. (outer wall data upshifted on u
+
axis by 1 wall unit). ( ) Water (inner),
Re =9900, v
+
max
= 0.83; ( ) water (outer), Re =9900, v
+
max
= 0.92; ( ) water (inner),
Re =30,600, v
+
max
= 0.99; ( ) water (outer), Re =30,600, v
+
max
= 1.08; ( ) water
(inner), Re =61,400, v
+
max
= 1.07; ( ) water (outer), Re =61,400, v
+
max
= 1.18; ()
0.0124% XG (Inner), Re =10,600, v
+
max
= 0.70; () 0.0124% XG (outer), Re =10,600,
v
+
max
= 0.72; () 0.0124% XG (Inner), Re =30,300, v
+
max
= 0.85; () 0.0124% XG
(outer), Re =30,300, v
+
max
= 0.90; () 0.0124% XG (inner), Re =57,600, v
+
max
= 0.92;
() 0.0124% XG (outer), Re =57,600, v
+
max
= 1.06. (For interpretation of the refer-
ences to colour in this gure legend, the reader is referred to the web version of the
article.)
Normalizing the axial rms uctuation component with the fric-
tion velocity, as done in Fig. 23, highlights an interesting effect
(note: the outer wall data in Figs. 2326 is upshifted on the u
+
axis
byonewall unit toseparatethedata). Suppressions canbeobserved
for 0.0124% xanthan gum at all Reynolds numbers for data mea-
surednear the outer wall while a slight increase is seenonthe inner
wall peak values. For the inner wall peak values the differences are
small andessentiallywithintheexperimental uncertaintyalthough
the trend is consistent across all the Reynolds numbers. At even
higher levels of drag reduction (DR=42.3% for 0.07% xanthan gum
1370 A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372
Fig. 26. Tangential rms uctuation levels in wall coordinates for water and 0.0124%
XG (outer wall data upshifted on u
+
axis by 1 wall unit). ( ) Water (inner),
Re =9900, w
+
max
= 0.99; ( ) water (outer), Re =9900, w
+
max
= 1.20; ( ) water
(inner), Re =30,600, w
+
max
= 1.25; ( ) water (outer), Re =30,600, w
+
max
= 1.42;
( ) water (inner), Re =61,400, w
+
max
= 1.45; ( ) water (outer), Re =61,400, w
+
max
=
1.65; () 0.0124% XG (inner), Re =10,600, w
+
max
= 1.34; () 0.0124% XG (outer),
Re =10,600, w
+
max
= 1.38; () 0.0124% XG (inner), Re =30,300, w
+
max
= 1.44; ()
0.0124% XG (outer), Re =30,300, w
+
max
= 1.55; () 0.0124% XG (inner), Re =57,600,
w
+
max
= 1.33; () 0.0124% XG (outer), Re =57,600, w
+
max
= 1.76. (For interpretation
of the references to colour in this gure legend, the reader is referred to the web
version of the article.)
Fig. 27. Radial rms uctuationlevels inwall coordinates for water and0.07%XG. ( )
Water (inner), Re =30,600, v
+
max
= 0.99; ( ) water (outer), Re =30,600, v
+
max
= 1.08;
() 0.07% XG (inner), Re =28,700, v
+
max
= 0.63; () 0.07% XG (outer), Re =28,700,
v
+
max
= 0.66. (For interpretation of the references to colour in this gure legend, the
reader is referred to the web version of the article.)
in Fig. 24), an increase in the peak levels is observed at both walls.
The radial rms uctuation levels for 0.0124% xanthan gum, shown
in Fig. 25, are globally much lower than that for water. In Fig. 26
increases in the peak values are observed for the tangential com-
ponent except for the peak value at the inner wall for Re =57,600.
For higher levels of drag reduction, as in the case for 0.07% xanthan
gum, increased suppression could be seen for both the radial and
tangential rms uctuation levels when plotted in wall coordinates
as shown in Figs. 27 and 28.
In Fig. 29 the peak values of the turbulent uctuation compo-
nents close to the inner and outer walls, normalized with the bulk
Fig. 28. Tangential rms uctuation levels in wall coordinates for water and 0.07%
XG. ( ) Water (inner), Re =30,600, w
+
max
= 1.25; ( ) water (outer), Re =30,600,
w
+
max
= 1.42; () 0.07% XG (inner), Re =28,700, w
+
max
= 1.13; () 0.07% XG (outer),
Re =28,700, w
+
max
= 1.18. (For interpretationof the references tocolour inthis gure
legend, the reader is referred to the web version of the article.)
Fig. 29. Peaks of axial, radial and tangential uctuation components normalized
with the bulk velocity, UB, plotted against drag reduction () current study, ()
CMC [15], () CMC, XG, LAPONITE/CMC [24]. Red: u

/UB; green: v

]U8; blue: w

]U8;
hollow symbols: inner wall; lled symbols: outer wall). (For interpretation of the
references to colour in this gure legend, the reader is referred to the web version
of the article.)
velocity, have been plotted against level of drag reduction together
with the available data from the literature for the same annulus
radius ratio of k 0.5 [15,24]. The lines in the gure are included to
guide the readers eye where clear trends are apparent. The limited
data from the current study for the semi-rigid polymer, xanthan
gum, and the data obtained from the literature show a decreas-
ing trend below 40% drag reduction of the normalized axial peak
level above which a slightly more complex but generally increas-
ing trend is observed. A vertical dotted line is also included in the
gure to demarcate these two regions. Apart froma slight increase
of the normalized tangential component for DR12%, decreasing
normalized radial and tangential components with drag reduction
can also be seen. Below40% DR the trends are very similar to those
identied by Escudier et al. [5] for ow through a rectangular duct
and by Japper-Jaafar et al. [7] for pipe ow. However above 40% DR
the trend in the current axial peak levels is different to the rect-
angular duct and pipe ow data. This discrepancy may be due to a
A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372 1371
lack of data for highly exible high molecular weight polymers (e.g.
polyethylene oxide or polyacrylamide) in the annular ow case.
Solutions of exible polymer molecules have beenshowntoexhibit
a different turbulent structure compared to more rigid molecules
such as xanthan in complex ows for example [57].
7. Conclusions
Laminar-turbulent transition for ow of Newtonian and non-
Newtonian liquids through an annular pipe was detected by
monitoring the axial rms uctuation level at xed radial locations
close to the inner and outer walls ( =0.1 and 0.9). It was found that
theReynolds-number rangefor transitional owwas greater for the
moreshear-thinninguids. Timetraces of themeanaxial velocityat
these radial locations providedsome insight intothe nature of tran-
sition within the ow. These observations indicate that the higher
shear stress on the inner wall compared to that on the outer wall
does not leadto earlier transitionfor shear thinning andyieldstress
uids as is observed for the Newtonian uids. Though velocity vari-
ations are detected within the time traces for the owcloser to the
inner wall, this effect could be a consequence of the high ampli-
tude spikes occurring for the owcloser to the outer wall, i.e. mass
conservation enforcing changes in velocity near the inner wall. The
mean axial velocity indicates a slightly different behaviour for the
Newtonian and polymer ow in relation to the location of max-
imum velocity. A shift towards the outer wall from a location of
=0.44 could be seen within the transitional ow regime for the
glycerinewater mixture due to the higher number of turbulence
spots which results in further increase of the shear stress in the
inner wall region compared to that in the outer wall region. The
modication is a consequence of the ow adjusting to the change
in momentumtransport as suggested by Hanks and Bonner [13]. In
contrast for the polymer ows such a shift in the location of max-
imum velocity is not detected and, if any, is not signicant due to
attening of the velocity proles associated with shear-thinning.
A qualitative analysis of the peak values of the turbulent uc-
tuation levels (normalized with U
B
) in the drag-reduction study
shows a decreasing trend of the axial component below 40% drag
reduction. Above this drag-reduction limit, the peak axial lev-
els increase, generally, with drag reduction in contrast to what
is observed in pipe and channel-ow studies [5,7]. Apart from a
slight increase of the normalized tangential uctuation component
for DR12%, the normalized radial and tangential rms uctuating
components remained below the Newtonian values for all drag-
reducingows, withthepeakvaluedecreasingwithincreasingdrag
reduction, similar to the pipe and channel-ow studies.
References
[1] R.R. Rothfus, C.C. Monrad, V.E. Senecal, Velocity distributionanduidfrictionin
smooth concentric annuli, Industrial and Engineering Chemistry 42 (12) (1950)
25112520.
[2] R.P. Chhabra, J.F. Richardson, Non-Newtonian Flow and Applied Rheology, 2nd
ed., Butterworth-Heinemann, Oxford, 2008, ISBN 978-0-7506-8532-0.
[3] F.A. Seyer, A.B. Metzner, Turbulence phenomenon in drag reducing system,
AIChE Journal 15 (3) (1969) 426434.
[4] F.T. Pinho, J.H. Whitelaw, Flow of non-Newtonian uids in a pipe, Journal of
Non-Newtonian Fluid Mechanics 34 (1990) 129144.
[5] M.P. Escudier, A.K. Nickson, R.J. Poole, Turbulent ow of viscoelastic shear-
thinning liquids through a rectangular duct: quantication of turbulence
anisotropy, Journal of Non-Newtonian Fluid Mechanics 160 (2009) 210.
[6] M.P. Escudier, S. Rosa, R.J. Poole, Asymmetry in transitional pipe ow of drag-
reducing polymer solutions, Journal of Non-Newtonian Fluid Mechanics 161
(13) (2009) 1929.
[7] A. Japper-Jaafar, M.P. Escudier, R.J. Poole, Turbulent pipeowof a dragreducing,
rigid rod-like polymer solution, Journal of Non-Newtonian Fluid Mechanics
161 (2009) 8693.
[8] M.P. Escudier, F. Presti, Pipe ow of a thixotropic liquid, Journal of Non-
Newtonian Fluid Mechanics 62 (1996) 291306.
[9] M.P. Escudier, R.J. Poole, F. Presti, C. Dales, C. Nouar, C. Desaubry, L. Graham,
L. Pullum, Observations of asymmetrical ow behaviour in transitional pipe
owof yield-stress andother shear-thinningliquids, Journal of Non-Newtonian
Fluid Mechanics 127 (2005) 143155.
[10] J. Peixinho, C. Nouar, C. Desaubry, B. Theron, Laminar transitional and turbulent
ow of yield stress uid in a pipe, Journal of Non-Newtonian Fluid Mechanics
128 (23) (2005) 172184.
[11] B. Gzel, I. Frigaard, D.M. Martinez, Predicting laminar-turbulent transition in
Poiseuille pipe ow for non-Newtonian uids, Chemical Engineering Science
64 (2009) 254264.
[12] B. Gzel, T. Burghelea, I. Frigaard, D.M. Martinez, Observations of laminar-
turbulent transition of a yield stress uid in HagenPoiseuille ow, Journal
of Fluid Mechanics 627 (2009) 97128.
[13] R.W. Hanks, W.F. Bonner, Transitional ow phenomena in concentric annuli,
Industrial and Engineering Chemistry Fundamentals 10 (1) (1971) 105113.
[14] K. Rehme, Turbulent ow in smooth concentric annuli with small radius ratio,
Journal of Fluid Mechanics 64 (2) (1974) 263287.
[15] J.M. Nouri, H. Umur, J.H. Whitelaw, Flow of Newtonian and non-Newtonian
uids in concentric and eccentric annuli, Journal of Fluid Mechanics 253 (1993)
617641.
[16] S.Y. Chung, G.H. Rhee, H.J. Sung, Direct numerical simulation of turbulent con-
centric annular pipe ow, Part 1: Flow eld, International Journal of Heat and
Fluid Flow 23 (2002) 426440.
[17] J.G. Knudsen, D.L. Katz, Fluid Dynamics and Heat Transfer, The McGraw-Hill
Companies, 1958, ISBN 0882759175.
[18] R.R. Rothfus, C.C. Monrad, K.G. Sikchi, W.J. Heideger, Isothermal skin friction
in ow through annular sections, Industrial and Engineering Chemistry 47 (5)
(1955) 913918.
[19] J.E. Walker, R.R. Rothfus, Transitional velocity patterns in a smooth concentric
annulus, AIChE Journal 5 (1) (1958) 5154.
[20] J.A. Brighton, J.B. Jones, Fully developed turbulent ow in annuli, Journal of
Basic Engineering 86 (1964) 835.
[21] C.J. Lawn, C.J. Elliott, Fully developed turbulent owthrough concentric annuli,
Journal of Mechanical Engineering Science 14 (3) (1972) 195204.
[22] K. Rehme, Turbulence measurements in smooth concentric annuli with small
radius ratios, Journal of Fluid Mechanics 72 (1) (1975) 189206.
[23] M.P. Escudier, P.J. Oliveira, F.T. Pinho, S. Smith, Fully developed laminar owof
non-Newtonian liquids through annuli: comparison of numerical calculations
and experiments, Experiments in Fluids 33 (2002) 101111.
[24] M.P. Escudier, I.W. Gouldson, D.M. Jones, Flow of shear-thinning uids in a
concentric annulus, Experiments in Fluids 18 (1995) 225238.
[25] I.H. Gucuyener, T. Mehmetoglu, Characterization of ow regime in concentric
annuli and pipes for yield-pseudoplastic uids, Journal of Petroleum Science
and Engineering 16 (1996) 4560.
[26] I.M. Mishra, P. Mishra, Transition from laminar to turbulent ow of purely vis-
cous non-Newtonianuids inconcentric annuli, IndianChemical Engineer XXII
(4) (1980) 3941.
[27] P. Mishra, G. Tripathi, Transition from laminar to turbulent ow of purely vis-
cous non-Newtonian uids in tubes, Chemical Engineering Science 26 (1971)
915921.
[28] R.J. Poole, Development length requirements for fully-developed laminar ow
inconcentric annuli, ASMEJournal of Fluids Engineering132(6) (2009) 064501.
[29] F.M. White, Viscous Fluid Flow, The McGraw-Hill Companies, 2005, ISBN 0-
071-24493-X.
[30] B.R. Munson, D.F. Young, T.H. Okiishi, Fundamentals of Fluid Mechanics, 4th
ed., John Wiley and Sons, Inc., 2002, ISBN 0-471-44250-X.
[31] F. Presti, Investigationof transitional andturbulent pipeowof non-Newtonian
uids, Ph.D. Thesis, University of Liverpool, 2000.
[32] D. Poggi, A. Porporato, L. Ridol, An experimental contribution to near-wall
measurements by means of a special laser Doppler anemometry technique,
Experiments in Fluids 32 (2002) 366375.
[33] C. Tropea, Laser Doppler anemometry: recent developments and future chal-
lenges, Measurement Science and Technology 6 (1995) 615619.
[34] W.J. Yanta, R.A. Smith, Measurements of turbulence-transport properties with
a laser Doppler velocimeter, in: AIAA 11th Aerospace Science Meeting, 1973.
[35] K. Yasuda, R.C. Armstrong, R.E. Cohen, Shear ow properties of concentrated
solutions of linear and star branched polystyrenes, Rheologica Acta 20 (1981)
163178.
[36] M.P. Escudier, I.W. Gouldson, A.S. Pereira, F.T. Pinho, R.J. Poole, Onreproducibil-
ity of the rheology of shear thinning liquids, Journal of Non-Newtonian Fluid
Mechanics 97 (2001) 99124.
[37] M. Milas, M. Rinaudo, M. Knipper, J.L. Schuppiser, Flow and viscoelastic prop-
erties of xanthan gum solutions, Macromolecules 23 (1990) 25062511.
[38] R. Lapasin, S. Pricl, Rheology of industrial Polysaccharides: Theory and Appli-
cations, Blackie Academic and Professional, 1995, ISBN 0 7514 0211 7.
[39] A.B. Rodd, D.E. Dunstan, D.V. Boger, Characterisation of xanthan gumsolutions
using dynamic light scattering and rheology, Carbohydrate Polymers 42 (2000)
159174.
[40] L.E. Rodd, T.P. Scott, J.J. Cooper-White, G.H. McKinley, Capillary break-
up rheometry of low-viscosity elastic uids, Applied Rheology 15 (2005)
1227.
[41] M. Stelter, G. Brenn, Elongational rheometry for the characterization of vis-
coelastic uid, Chemical Engineering and Technology 25 (1) (2002) 3035.
[42] E. Pelletier, C. Viebke, J. Meadows, P.A. Williams, A rheological study of the
order-disorder conformational transition of xanthan gum, Biopolymers 59
(2001) 339346.
[43] G.P. Roberts, H.A. Barnes, New measurements of the ow-curves for Carbopol
dispersions without slip artefacts, Rheologica Acta 40 (2001) 499503.
1372 A. Japper-Jaafar et al. / J. Non-Newtonian Fluid Mech. 165 (2010) 13571372
[44] H. Zhu, Y.D. Kim, D.D. Kee, Non-Newtonian uids with a yield stress, Journal of
Non-Newtonian Fluid Mechanics 129 (2005) 177181.
[45] R.K. Shah, A.L. London, Laminar Flow Forced Convection in Ducts, Academic
Press, New York, 1978, ISBN 0-12-020051-1.
[46] O.C. Jones, J.C.M. Leung, Animprovement inthe calculationof turbulent friction
insmoothconcentric annuli, Journal of Fluids Engineering 103(1981) 615623.
[47] J.L. Zakin, C.C. Ni, R.J. Hansen, Laser Doppler velocimetry studies of early tur-
bulence, Physics of Fluids 20 (102) (1977) 8589.
[48] J.T. Park, R.J. Mannheimer, T.A. Grimley, T.B. Morrow, Pipe ow measure-
ments of a transparent non-Newtonian slurry, Journal of Fluids Engineering
111 (1989) 331336.
[49] I.J. Wygnanski, F.H. Champagne, On transition in a pipe. Part 1. The origin of
puffs and slugs and the ow in a turbulent slug, Journal of Fluid Mechanics 59
(2) (1973) 281335.
[50] Y. Rubin, I.J. Wygnanski, J.H. Haritonidis, Further observations on transition
in a pipe, in: R. Eppler, H. Fasel (Eds.), Laminar-Turbulent Transition. IUTAM
Symposium Stuttgard, Germany, 1979, pp. 1726.
[51] D.H. Zhang, Y.T. Chew, S.H. Winoto, Investigation of intermittency measure-
ment methods for transitional boundary layer ows, Experimental Thermal
and Fluid Sciences 12 (1996) 433443.
[52] A. Jaafar, Duct ow of polymer solutions, Ph.D. Thesis, University of Liverpool,
2010.
[53] F.H. Clauser, The turbulent boundary layer, Advances in Applied Mechanics 4
(1956) 151.
[54] H. Tennekes, J.L. Lumley, A First Course in Turbulence, The MIT Press, 1972,
ISBN 978-0-262-20019-6.
[55] M.D. Warholic, H. Massah, T.J. Hanratty, Inuence of drag-
reducing polymers on turbulence: effects of Reynolds number,
concentration and mixing, Experiments in Fluids 27 (1999)
461472.
[56] M.P. Escudier, F. Presti, S. Smith, Drag reduction in turbulent pipe
ow of polymers, Journal of Non-Newtonian Fluid Mechanics 81 (1999)
197213.
[57] R.J. Poole, M.P. Escudier, Turbulent ow of viscoelastic liquids through an
axisymmetric sudden expansion, Journal of Non-Newtonian Fluid Mechanics
117 (2004) 2546.